Signal Processing
Spiking Neural Network-based Equalization and Demapping in IM/DD Systems: A Comparison
Eike-Manuel Edelmann, Alexander von Bank and Laurent Schmalen
In recent years, artificial neural networks (ANNs) have emerged as the de facto standard for addressing challenging problems in communications engineering that are difficult to solve using traditional methods. However, the dense
matrix multiplications employed in ANNs often result in high computational complexity and, consequently, power-hungry systems. To overcome these limitations, researchers are turning to alternative computational models, such as spiking neural networks (SNNs), which promise highly energy-efficient computation. Recent works have shown that SNN-based equalizers and demappers achieve promising results in non-coherent short-reach optical communication systems affected by chromatic dispersion and nonlinear distortion. However, the proposed approaches differ in several design choices, such as the incorporation of decision feedback, the use of recurrent connections inside the SNN, different neural encodings, and regularization techniques. A systematic comparison of the various design choices has not yet been reported in the literature. This article presents a systematic comparison of design choices for SNN-based equalizers and demappers in intensity modulation with direct detection systems, evaluating their impact on bit error rate, spike activity, and model size. The results reveal that design and encoding choices affect these metrics differently, leading to inherent trade-offs.
Read the full abstract ▾
General Relativity and Quantum Cosmology
On the quasiblack-hole limit of rotating charged fluids
Marcos L. W. Basso and Vilson T. Zanchin
We investigate the extremal quasiblack-hole (QBH) limit of stationary, axisymmetric charged perfect fluids in rigid or differential rotation, with and without pressure. We emphasize Weyl-type configurations, whose
redshift factor is functionally related to the generalized electromagnetic potential. In this limit, the
redshift factor vanishes throughout the fluid interior and the boundary becomes a quasihorizon. We require regular matter and electromagnetic fields and smooth matching to the exterior. Under suitable convergence assumptions, electromagnetic regularity and the approach to uniform rotation imply a constant generalized electromagnetic potential throughout the connected fluid interior, independently of the Weyl ansatz. With additional integrability conditions, the mass formula reduces to the extremal Kerr-Newman Smarr relation. For rigid rotation, we examine charged
dust obeying a linear Weyl relation and fluids with pressure obeying the Kloster-Das or Guilfoyle relations. The linear Kloster-Das subclass becomes pressureless in the limit, whereas the general Guilfoyle case allows nonzero pressure. The Islam ansatz obstructs a regular limit when its coupling parameter, limiting potential, and limiting charge density are nonzero. For differential rotation, we analyze configurations with an identically vanishing Lorentz-force term and a linear Weyl subclass whose regularity requires control of angular-velocity gradients. Our results show that rotating Weyl-type systems admit extremal QBH limits much like their static counterparts, extending analyses of rotating
dust distributions and identifying conditions for more general rotating charged fluids to be compatible with this limit.
Read the full abstract ▾
Computer Vision
One Geometry, Different Outcomes: Readout-Dependent Effects of the Modality Gap in Vision-Language Models
Aditya Sharma and Divya Saxena
Contrastive
vision-
language models learn shared
embedding spaces by aligning matched image-text pairs, yet their representations remain separated by a modality gap. Prior work reports divergent effects of modifying this gap: reducing it can improve zero-shot classification and cross-modal
alignment, whereas removing gap-related structure can degrade image-text
retrieval. In this paper, we provide a unified
geometric explanation for these task-dependent effects. Across CLIP and SigLIP encoders, we find that a single dominant direction captures 94.4-99.9% of the squared norm of the image-text mean separation, revealing that the mean-separation component is approximately rank-one. A decomposition of the similarity score then identifies three task-specific roles. In zero-shot classification, query-side fixed gap-offset subtraction is exactly equivalent to an additive class bias. In standard cross-modal
retrieval, projecting out the gap direction and renormalising residuals discards candidate-specific norm information, inducing a multiplicative ranking distortion; a geometry-derived exponent tracks the grid-search optimum (Spearman rho = 0.93) and restores performance in some settings, although the gains transfer unevenly. In mixed-modal
retrieval, the gap direction sorts candidates by modality; its removal can improve cross-modal ranking, unlike random or non-gap controls. Residual semantic structure after removal defines the limits of the rank-one account. Together, these results explain why gap modification can improve, degrade, or restore performance across downstream settings. By clarifying when and why gap modification changes model behavior, this account provides a principled basis for selecting gap interventions in similarity-based
vision-
language systems across evaluated downstream tasks.
Read the full abstract ▾
Materials Science
Accelerating phase-field simulations on exascale computing systems for faster-than-real-time precipitate aging predictions
Stephen DeWitt, David J. Gardner, Philip Fackler et al. (+4 more)
Three-dimensional phase-field simulations are a gold-standard for microstructure prediction for materials, but their computational cost often limits application-relevant calculations to modest domain sizes and timescales. Here, we present a holistic approach for accelerating
Fourier pseudospectral phase-field simulations by combining performance-portable GPU computing, large-scale distributed-
memory parallelism, and high-order implicit-explicit time integration within the MEUMAPPS C++ framework. We demonstrate this approach with a 1.5-billion-grid-point simulation of the growth and coarsening of 1,920 gamma” precipitates in a Ni-Nb-Fe
alloy. The simulated seven-hour heat treatment is completed in 5.5 hours, making the calculation faster than real time and an estimated 217-501x faster than a CPU-only, first-order baseline. This performance makes three-dimensional simulations of thousands of interacting precipitates tractable, enabling quantitative studies ocollective microstructural phenomena, large simulation ensembles, and real-time integration into controlling and interpreting experiments. Benchmarking shows single-node GPU speedups of up to 17.5x relative to comparable CPU resources, near-ideal strong scaling to 4096 GPUs for a 12-billion-grid-point problem, and a further 2.6-2.9x acceleration from fourth-order time integration at scientifically relevant error tolerances. These results establish a performance-portable strategy for exploiting leadership-scale GPU systems that is applicable to a broad class of
Fourier pseudospectral simulations beyond phase-field models such as fluid dynamics and
crystal plasticity simulations.
Read the full abstract ▾
All authors Stephen DeWitt, David J. Gardner, Philip Fackler, Yonggil Song, Miroslav Stoyanov, Carol S. Woodward, Balasubramaniam Radhakrishnan
Machine Learning
Data Unlearning via Inverse Distillation
Aleksei Leonov, Nikita Kornilov, Zhenhe Zhang et al. (+3 more)
Multi-step matching models, including flow and
diffusion models, produce high-quality outputs but incur substantial inference costs and may reproduce unwanted components of their training
datasets. We introduce Inverse
Distillation Unlearning (IDU), a unified framework that simultaneously distills a teacher multi-step matching model into an efficient one-step student generator and suppresses outputs corresponding to a designated training subset. We first formulate
distillation as a min-max objective over a data distribution and then represent this distribution as a mixture of the forget-set and the generated distributions. This allows us to compare this mixture with the teacher's training distribution and recover only the retained data at the optimum. Our method requires only a pretrained full-data teacher and data from the forget set, without access to retained training examples, extra feature extractors or classifiers. Extensive experiments on MNIST and CIFAR-10
datasets under flow-matching and score-based
diffusion settings demonstrate that IDU substantially reduces the generation frequency of forgotten classes while preserving generation quality on the retained classes. To the best of our knowledge, IDU is the first unified framework for simultaneous unlearning and
distillation in unconditional flow-matching and score-based models.
Read the full abstract ▾
All authors Aleksei Leonov, Nikita Kornilov, Zhenhe Zhang, Evgeny Burnaev, Iaroslav Koshelev, Alexander Korotin
Group Theory
Small Kazhdan constants for small generating sets of linear groups
Tsachik Gelander, Alex Lubotzky and Jvbin Yao
Let G be a linear group with Kazhdan property (T) generated by d elements. We show that there are generating sets of size d+2 with arbitrarily small Kazhdan constants. We also prove the same conclusion for every finitely generated group admitting a non-elementary convergence action on a compact metrizable space. This includes non-elementary word-
hyperbolic and relatively
hyperbolic groups, as well as finitely generated non-elementary subgroups of word-
hyperbolic groups.
Read the full abstract ▾
Neurons and Cognition
Better Behavioral Prediction, More Faithful Model Ablations? Evidence from Sequential Choice
Hanbo Xie
Using predictive models to explain cognition requires more than accurate behavioral predictions. Input ablations offer an appealing route: remove information from a model and interpret the resulting performance change as evidence of its importance for behavior. Yet this inference assumes that the model's dependence on information reflects the dependence of the process generating the behavior. We test it in two synthetic sequential bandit tasks with known generating policies, where past choices can remain informative when feedback is unavailable to a predictor. We compare GRUs and
Transformers trained from scratch, a fine-tuned LLaMA model, and cognitive models across systematically varied reward contributions. Our analyses distinguish prediction after training without reward observations from the response of a fixed predictor to donor-reward replacement. Three findings emerge. First, in the restless task, neural models trained without rewards predict held-out choices better than four simple training-fitted behavioral baselines. Second, under matched donor replacement, accurate predictors can respond much less than the known generator. Third, at some reward weights, neural networks predict better than a pooled reinforcement-learning model but have less faithful changes in choice probabilities; the model ordering differs between the two tasks. These independent-test results separate information sufficient for prediction from response fidelity under a specified ablation in sequential choice. They motivate validating model-ablation responses independently of predictive performance before using them to infer how the observed behavior was generated.
Read the full abstract ▾
Instrumentation and Methods for Astrophysics
Comparing Active Anomaly Detection Frameworks: From Light Curves to Images
Sreevarsha Sreejith and Robert C. Nichol
The increasing volume and complexity of data from modern astronomical surveys are making comprehensive visual inspection increasingly impractical, motivating automated methods for identifying rare and unusual objects. Active anomaly detection (A-AD) combines unsupervised anomaly detection with iterative user feedback, allowing searches to adapt to scientific interests while retaining sensitivity to unexpected objects. We systematically compare three A-AD frameworks, Astronomaly, Protege, and PineForest, across Zwicky Transient Facility light curves and Euclid
galaxy images. In baseline experiments, Protege, PineForest, and Astronomaly recovered 83, 78, and 38 variable
stars, respectively, within the first 100 light-curve objects inspected, while PineForest recovered 70 merging or morphologically disturbed
galaxies within the first 100 Euclid images, compared with 29 and 19 for Astronomaly and Protege. Experiments using prior labels show that their effectiveness depends on the composition of the prior sample and varies between A-AD methods. Finally, we combine the complementary behaviour of Protege and PineForest in a hybrid search of 10,000 Euclid images, recovering 799 morphologically unusual objects. These include 29 retained strong lens systems and candidates, of which 9 candidates have no counterparts in the published Euclid Q1 strong lens catalogues searched, and 202 visually classified ring
galaxies. These results highlight the importance of selecting A-AD strategies according to both the scientific objective and
dataset properties, and demonstrate their potential for efficient, discovery-driven exploration of forthcoming astronomical surveys.
Read the full abstract ▾
Differential Geometry
A Note on the Classification of Centroaffine Hypersurfaces
Zuzanna Szancer
Cao and Wang recently classified a class of locally strongly convex centroaffine hypersurfaces whose non-vanishing Tchebychev vector field $T$ satisfies $K(T,X)=\alpha X$ and $X(\alpha)=0$ for every $X\perp T$, together with $K(T,T)=\beta T$. In the degenerate case $\beta=2\alpha$, their higher-dimensional argument assumes in addition that the $T^\perp$-component of the difference tensor $K$ vanishes identically. We show that this extra assumption is unnecessary and the
theorem can be strengthened.
Read the full abstract ▾
Systems and Control
Emergency Control of Transmission Voltages by Coordinating DERs From Multiple Substations
Francisco Escobar and Gustavo Valverde
Voltage stability can be endangered in transmission networks that operate under high strain and exhibit a large separation between generators and loads. During emergencies associated to voltage instability, one promising supporting countermeasure is to control large populations of Distributed Energy Resources (DERs) connected to medium- and low-voltage networks, so that they locally reduce the net load by modifying their power generation or consumption. However, it is challenging to achieve this control while anticipating interactions with legacy voltage regulators, coordinating efforts across substations, and exchanging small amounts of information with DERs. In this paper, we propose a control scheme that coordinates DER aggregations with the load tap changers of substation
transformers, asks them to contribute power depending on the substation they are connected to, and relies on lightweight communication. Our simulations show that the proposed control can successfully coordinate DERs and load tap changers to enhance long-term voltage stability.
Read the full abstract ▾
Theoretical Economics
Endogenous Information Design and Informational Evolution in Dynamic Economic Systems with Incentives and Learning
Guoqiang Tian
This paper studies the endogenous informational
evolution of dynamic economic systems. Building on Hurwicz's conceptual framework, such a system is defined as a dynamic joint adjustment process of resource and information allocation linked by recursive feedback. Economic theory gains precision by isolating particular margins, but when information production, disclosure, learning, incentives, actions, and future feasibility interact across these two processes, conclusions obtained by holding other margins fixed need not survive their joint determination. The paper develops a system-level dynamic
equilibrium framework for this interaction. Holding particular links fixed yields familiar environments in dynamic mechanism design, information design, experimentation and learning, feedback and disclosure, and information acquisition. The analysis establishes three central failures relative to fixed or partially exogenous informational environments: informational efficiency under exogenous feasibility is generally incompatible with endogenous dynamic feasibility; incentive and disclosure design cannot generally be separated; and informational monotonicity can fail without implementation replicability. It also identifies the corresponding positive boundaries. Optimal disclosure may be interior or extreme; bang–bang disclosure arises under additional conditions, and a further specialization generates an endogenous transition from opacity to transparency. Information may therefore be optimally withheld even when its direct decision value is strictly positive. Extensions to private payoff-relevant information and limited commitment preserve the recursive structure.
Read the full abstract ▾
Cosmology and Nongalactic Astrophysics
Constraining reionization with persistent homology: a novel summary statistic for wide-field Lyman-$\alpha$ emitter observations
Michael M. Wyatt, Nikolaos Triantafyllou, Steven R. Furlanetto et al. (+4 more)
The spatial clustering of Lyman-$\alpha$ emitters (LAEs) is one of the most promising observables for constraining reionization. In this work, we introduce persistent homology, a popular tool in topological data analysis, and specifically the Betti-0 curve as a summary statistic to characterize the clustering of LAEs. We compare the performance of the Betti-0 curve against the angular correlation function (ACF) and void probability function (VPF) by quantifying their ability to constrain the global neutral fraction $x_{\rm HI}$ during reionization using the Fisher information. We forward model observations of LAEs using 21cmFAST by including $\mathrm{Ly}\alpha$ emission as a post-processing step and account for
cosmic variance uncertainties. We focus our analysis within the framework of the SILVERRUSH. catalog at $z=5.7$ and 6.6, and find that the Betti-0 curve performs as well as or better than the ACF and VPF. We also find that the statistics are most sensitive to different physical scales and thus probe complementary information, and as a consequence that simultaneously constraining all three statistics improves constraints by a factor of $\sim 2$ compared with the ACF alone, with predicted minimum uncertainties of $\sigma_{\rm min} \sim$ 0.03 and 0.07 at $x_{\rm HI} = 0.05$ and $z=5.7$ and $x_{\rm HI} = 0.20$ and $z=6.6$, respectively. Our results show that the Betti-0 curve is a promising summary statistic to constrain reionization using LAE clustering and motivate further exploration of persistent homology for the same purpose in other wide-field surveys such as those made possible by the Nancy Grace Roman
Telescope.
Read the full abstract ▾
All authors Michael M. Wyatt, Nikolaos Triantafyllou, Steven R. Furlanetto, Hiroya Umeda, Samuel Gagnon-Hartman, Andrei Mesinger, Masami Ouchi
Robotics
MagNav: A Dual-Core Magnetic Track Guidance Framework for Lighting-Invariant Navigation in Two-Wheeled Robots
Arpita Kumari, Anupam Chatterjee and Richa Srivastava
Two-Wheeled Inverted Pendulum (TWIP)
robots are useful for studying how to control systems that are naturally unstable and have fewer actuators than degrees of freedom. Adding autonomous line-following to these
robots is challenging because steering and balancing are closely linked. Most existing systems use infrared sensors, which can be affected by changes in lighting, such as sunlight or shadows, making them reliable only indoors. This paper presents a self-balancing
robot that can follow a line using a
magnetic track guidance system. By using a five-channel analog
Hall-effect sensor array, the
robot is not affected by optical interference. The control system uses a cascaded PID structure: the inner loop keeps the
robot balanced using data from an inertial measurement unit with a complementary filter, while the outer loop adjusts steering based on the
magnetic sensor readings. Stepper motors provide precise torque control without needing extra rotary encoders. For comparison, an optical sensor module was also included. Tests show that the
magnetic guidance system keeps accurate tracking even in very bright lighting, over 10,000 Lux, while the optical system loses accuracy and sometimes fails. This design provides a reliable, lighting-independent solution for autonomous
navigation in places like factories, warehouses, and outdoor paths.
Read the full abstract ▾
Systems and Control
Control and Estimation Co-Design via Envelope-Theorem Gradients
Mohammad S. Ramadan, Philip Dinenis and Mihai Anitescu
Instead of the sequential plant–control–
estimator design pipeline, which is in general not optimal, we propose the Control and Estimation Co-Design (ContEst) framework, which poses the entire system design as a single two-stage problem solved by first-order methods. A key enabler is that the gradient of each inner control/estimation optimal cost with respect to the design parameters is available from the dual variables the inner solver already returns, by the envelope
theorem: exactly under mild regularity (a unique inner minimizer), and as a subgradient under more relaxed conditions. That is, no differentiation through the optimizer or its optimality conditions is needed. We first present the framework in a linear quadratic Gaussian (LQG) regime, where the control and estimation costs become two semidefinite programs (SDPs) whose design gradients are read from the duals of the Lyapunov constraints. We extend the results to robust ($H_\infty$) formulation to cover worst-case control and filtering problems. We also present an extension to constrained and nonlinear systems, using an extended Kalman filter (eKF) and a local linearization of the information-state dynamics to yield an approximate but convex model predictive control (MPC) problem whose design gradients are read directly from the costates. We apply our methods to a variety of realistic design examples from different fields and report significant design improvements.
Read the full abstract ▾
Commutative Algebra
A note on inverse limits of injective modules
Chencheng Zhang
Let $R$ be a ring and $M$ a left $R$-module. We prove that $M$ is an $\omega_1$-indexed inverse limit of injective modules with split epimorphic connecting maps, and $M$ is also an $\omega$-indexed inverse limit of a stationary tower defined by a split monic endomorphism of one injective module. These constructions answer two questions of Bergman. We then study the class of direct summands of $\omega$-indexed inverse limit of injective modules with epimorphic connecting maps. Over a commutative Noetherian ring $R$, they are precisely the Enochs cotorsion modules if and only if $\operatorname{gldim}R\leq1$ and $\operatorname{Spec}R$ is countable.
Read the full abstract ▾
Combinatorics
Resultants, Recursive Formulas and Binet Formulas for $\textrm{SU}(N)$ Verlinde Sums
Jay Jorgenson, Anders Karlsson and Lejla Smajlović
We study Verlinde sums $V_{n}(N,m)$, which for our purposes are finite sums of trigonometric functions $\{r_{x}\}$, associated to $\textrm{ SU}(N)$. In this article we prove the following results: (i) the
polynomial $D_N(w):=\prod_x(w-r_x)$ is equal to, up to an explicit and computable factor, the iterated resultant of $Q(u)=u^m-1$ with an explicit elementary
polynomial $W_N$ which is independent of $n$ and $m$; (ii) for fixed $N$ and $m$, $V_n(N,m)$ satisfies a linear recursion of order at most $\binom{m-1}{N-1}$, together with arguments showing that one actually has a recursion of length equal to the number of distinct roots of $D_{N}$; (iii) for fixed $N$ and $m$, a Binet-type closed form expressing $V_n(N,m)$ as an explicit finite sum of $n$-th powers of algebraic numbers which are rescaled reciprocals of the distinct roots of $D_N(w)$ is derived. Several explicit examples are provided for $N=2,3,4,5,6$ and $8$.
Read the full abstract ▾
Machine Learning
PHASE: A Physiology-Guided Hierarchical Foundation Model for Intracranial EEG
Yipeng Zhang, Chenda Duan, Yuanyi Ding et al. (+10 more)
Clinicians and neuroscientists have long analyzed intracranial electroencephalography (iEEG) through directly measurable physiological characteristics, which carry much of the information that downstream tasks depend on. Recent iEEG foundation models learn by reconstructing or predicting their inputs, which leaves the retention of these characteristics implicit. They are also evaluated mainly on cognitive decoding and a narrow clinical task, i.e., seizure detection. On a broad, clinically relevant
benchmark such as Omni-iEEG, they remain below task-specific models when used frozen. We introduce PHASE, a physiology-guided foundation model that makes these characteristics explicit learning targets, pairing them with masked latent prediction in a temporal stage (PHASE-T) within each channel and a spatiotemporal stage (PHASE-ST) across synchronized channels. PHASE is pretrained on heterogeneous recordings from 222 participants at nine clinical sites. On all five Omni-iEEG clinical tasks, frozen PHASE-T outperforms every evaluated foundation model by up to 31%, and fine-tuned PHASE-T surpasses the task-specific models, setting a new state of the art. PHASE-T benefits from physiological supervision, outperforming variants trained with latent prediction alone or auxiliary waveform reconstruction on every task in matched ablations. PHASE-T generalizes to unseen institutions, outperforming the compared models with few or no local labels. PHASE-ST further improves seizure-onset-zone identification over PHASE-T and, when frozen, decodes sound volume and pitch on BrainTreebank better than published models. Beyond task performance, PHASE learns to encapsulate the physiological characteristics clinicians recognize, from seizure onset and its propagation to anatomical region identity, even though its pretraining contains no ictal recordings or anatomical labels.
Read the full abstract ▾
All authors Yipeng Zhang, Chenda Duan, Yuanyi Ding, Tianyi Wang, Atsuro Daida, Masaki Izumi, Yuta Tanoue, Naoto Kuroda, Shaun A. Hussain, Nishant Sinha, Eishi Asano, Hiroki Nariai, Vwani Roychowdhury
Computation and Language
PADMÉ: Preference Alignment Data Synthesis for Meta-Evaluation of LM Agent Evaluators
Cheng Chang, Yining Mao and Peng Qi
Language models are frequently employed to evaluate other
language models. An LM evaluator scoring agentic behaviors across multiple criteria is valuable, provided that its decisions align with human judgment. We call the problem of evaluating this
alignment Meta-Evaluation. Tackling it directly is difficult: collecting human data is expensive, absolute scoring is hard to align, and using an LM meta-evaluator recurses the question of trustworthiness. We adopt a reformulation of meta-evaluation as a preference judgment problem: rather than comparing human and LM evaluator scores of a trajectory, we ask whether their implied preferences align. Building on this, we introduce PADMÉ, a data synthesis method that generates reliable criterion-based meta-evaluation data for agentic settings. PADMÉ uses only small
language models, requires no human involvement during evaluations, and operates under a low computational budget. We build a prototype of PADMÉ and synthesize a
dataset of 1,000 samples across four agentic domains and three evaluation criteria. Human validation on a 150-sample subset demonstrates that PADMÉ improves agreement with human judgment from 73% to 85% over a naive baseline. Meta-evaluating 25 common models with our
dataset demonstrates the correlations between evaluation performance and scoring granularity, leniency, and model size, among other factors.
Read the full abstract ▾
Plasma Physics
Electron Density Measurements from Stark Broadened Emission in a Sodium Plasma Produced by Laser Resonance Saturation
Mark A. Cappelli
Electron Stark broadening of the $4^2D$–$3^2P$ multiplet transition in a sodium
plasma produced by
laser resonance saturation has provided a means of undertaking the first spatial measurements of the free electron density across and along the
plasma channel created in sodium vapor of density $10^{15}$–$5\times10^{16}$~cm$^{-3}$. From these measurements and measurements of the neutral sodium density within a heat sandwich oven, we are able to deduce the corresponding electron temperature. These temperatures compare favourably with the electron temperature estimated from a Boltzmann analysis of line intensities and suggest that a highly ionized ($>$10%)
plasma of electron temperature $<$6000K can be produced within 100~ns of
laser excitation from a
laser of modest irradiance ($10^6$–$10^7$~Wcm$^{-2}$). The experimental results are in reasonable agreement with a recently developed 3-dimensional model which predicts a decrease in the degree of ionization along the path of the
laser beam as a result of significant depletion of
laser energy. These experimental results demonstrate that this LIBORS code is capable of predicting the 3-dimensional nature of this new mode of
laser ionization with reasonable accuracy, and may also explain the low electron temperatures and free electron densities observed by other research teams.
Read the full abstract ▾
Materials Science
Measuring trainable degrees of freedom in materials graph neural networks: a random-subspace intrinsic dimension analysis
Shehroz Ahmad Shoaib and Kangming Li
Final predictive accuracy is the standard basis for comparing graph neural networks (GNNs) in materials-property prediction, but it does not show how strongly performance depends on access to trainable parameter-space directions. Here, we introduce trainable-degree dependence as a complementary characterization of materials GNN learning. Using random-subspace intrinsic-dimension analysis, we train CGCNN, ALIGNN, and DimeNet++ in randomly oriented parameter subspaces across six prediction tasks and measure how performance recovers as independent trainable degrees of freedom are restored. The resulting recovery curves separate endpoint accuracy from the trainable-dimensional demand required to recover it. They reveal distinctions that final errors alone miss: metallic classification and log-bulk-modulus regression recover near-reference performance from small fractional subspaces, formation-energy and band-gap prediction show stronger architecture dependence, and
phonon prediction is most sensitive to dimensional restriction.
Dataset-size sweeps show that band-gap models require larger fractional subspaces as training data grows, whereas formation-energy and bulk-modulus responses are more stable. A width sweep shows that fractional thresholds can remain stable while absolute threshold dimensions increase with model size. Random-subspace analysis therefore provides a targeted stress test for how materials GNNs use their optimization space.
Read the full abstract ▾
Instrumentation and Detectors
Design and Analysis of a 2D Vernier Structure
Hüsna Yildiz and Rick Helmer
This paper presents the design and simulation of a 2D Vernier Time-to-Digital Converter (TDC) targeting less than 20 picosecond time resolution and 5-bit output resolution. The TDC comprises voltage-controlled delay buffers, NAND-based SR latches with reset functionality, and a calibration loop using a phase-frequency detector (PFD) and charge pump. The calibration loop dynamically adjusts delay mismatches to maintain consistent timing across process-voltage-temperature (PVT) corners, significantly improving linearity. Simulations show that the TDC achieves a resolution of 16.9 ps for tt corner and maintains DNL and INL below 0.5 LSB under all tested PVT conditions
Read the full abstract ▾
Artificial Intelligence
GeoOutageBench: Benchmarking Ambiguity-aware, Ontology-grounded Geospatiotemporal KGQA for Multimodal Power Outage and Resilience Analysis
Ethan D. Frakes, Amy Kvien, Rishabh Kundu et al. (+8 more)
We introduce GeoOutageBench, a
benchmark for assessing LLM-based geospatiotemporal KGQA for
multimodal outage and resilience analysis. Unlike existing KGQA
benchmarks for Web knowledge, GeoOutageBench considers a spatiotemporal KG that integrates visual, textual, and structured data from outage records, remote sensing, weather observations, storm and power events, geographic entities, and domain ontologies. It provides a competency query taxonomy at different difficulty levels from spatiotemporal containment and proximity, spatiotemporal co-occurrence analysis,
multimodal evidence, to hypothetical evaluation. Over
multimodal KG and query classes, GeoOutageBench provides user-configurable evaluation of three important, highly coherent yet less studied tasks: (1) LLMs' understanding for ambiguous geospatiotemporal questions in terms of NL to SPARQL interpretation, (2) query-driven assessment of ontology utility, and (3) answer accuracy of
multimodal KGQA
retrieval. GeoOutageBench provides a design principle and foundation for assessing LLM-KG systems that support real-world infrastructure resilience analysis. Our
benchmark, source code, data, results, and other documentation are available at https://github.com/UCF-SAGE/GeoOutageBench.
Read the full abstract ▾
All authors Ethan D. Frakes, Amy Kvien, Rishabh Kundu, Redad Mehdi, Van D. Tran, Vibha S. Mandayam, Kristopher O. Davis, Erika I. Barcelos, Roger H. French, Yinghui Wu, Mengjie Li
Machine Learning
Early Learning Shapes Later Directions Of Representation Change In Continual Learning
Yuantao Deng, Jinnuo Liu, Kaizhen Tan et al. (+1 more)
Representations continually change as a network learns new tasks. We ask whether early representational changes naturally form a
geometric structure that continues to shape later learning. We identify a low-dimensional subspace of early representation drift, which we call a scaffold, and test whether it is reused across subsequent tasks. Across four pretrained visual encoders and two
datasets, later representational changes consistently favor this early-defined subspace over matched random alternatives. This reuse is history-dependent: when networks experience different early tasks but identical later training inputs, each network preferentially reuses the scaffold induced by its own learning history. The same preference appears in individual optimizer updates, even though the network's dominant local response directions shift away from the original scaffold. Finally, constraining motion within the scaffold slows new-task acquisition more than matched random constraints, while effects on old-task retention are less consistent. In summary, these results suggest that early experience leaves a persistent
geometric imprint on how neural networks adapt to future tasks.Code is available at https://github.com/YuantaoDeng/latent-scaffold.
Read the full abstract ▾
All authors Yuantao Deng, Jinnuo Liu, Kaizhen Tan, Yuchen Liu
Quantum Physics
Witness-based quantification of non-Markovianity without complete Choi state reconstruction
Pritam Roy, Saheli Mukherjee and Bivas Mallick
Quantifying non-Markovianity is a fundamental problem in the theory of open
quantum systems, as it is essential for exploiting environmental
memory as a resource in
quantum technologies. Existing resource-theoretic and operational quantifiers, however, generally require complete knowledge of the intermediate dynamical map and are therefore experimentally demanding. Here, we develop a witness-based framework for the quantification of non-Markovianity that bypasses complete reconstruction of the Choi state associated with the intermediate dynamical map. We first show that every non-Markovianity witness provides a certifiable lower bound on the
geometric resource measure of non-Markovianity and identify the necessary and sufficient conditions under which this bound is tight. Motivated by the fact that non-positive-semidefinite witnesses are generally difficult to measure in real experimental scenario, we then focus on positive-semidefinite witnesses, which admit a natural implementation as positive operator valued measure (POVM) elements. We prove that every positive witness yields a certified lower bound on the Rivas–Huelga–Plenio (RHP) measure, thereby establishing a complete hierarchy connecting witness-based quantification, the RHP measure, and the
geometric distance measure. Applying our framework to amplitude-damping and general Pauli dynamics, we find that the optimal positive witnesses reduce to simple Bell-parity projectors. Furthermore, we show that the resulting witness values can be obtained directly from a small number of experimentally accessible correlation measurements, thereby eliminating the need for complete Choi-state reconstruction and providing an efficient and experimentally feasible approach for the quantification of non-Markovianity in open
quantum systems.
Read the full abstract ▾
Artificial Intelligence
A Polyphonic Conception of AI Understanding
Matthieu Queloz and Pierre Beckmann
When a doctor, a judge, or an engineer must decide whether to trust an AI model's output, they cannot avoid asking what the model understands. Purely mathematical or statistical descriptions struggle to distinguish trustworthy from untrustworthy outputs without reintroducing the question of AI understanding in all but name. Yet the question is ill-framed as it stands, because the inherited concept operates within a monophonic paradigm: the idea that a cognitive system's understanding of something must be localised to a single mechanism underpinning all the capacities conferred by such understanding. Drawing on a wide range of mechanistic evidence, we show that LLMs are pervasively polyphonic: outputs emerge from coalitions of parallel mechanisms of uneven reliability, which variously complement, duplicate, or drown out one another, with several coalitions sufficing for a task without any one being indispensable. Polyphony not only complicates attributions of understanding, but renders monophonic inference patterns hazardous. In response, we develop a conception of understanding fit for polyphonic AI. It centres on sound circuitry that is reliably and correctly recruited and in control of outputs. Attributions of understanding thereby become tractable claims about internal organisation, and can do the work of guiding trust in AI.
Read the full abstract ▾
Algebraic Geometry
Realizability of Cohomology Classes in Orthogonal Grassmannians
Naima Nader
Building on the work of Coskun and Ross, we investigate which
cohomology classes of the orthogonal Grassmannian $OG(k,n)$ can be realized as irreducible subvarieties of $OG(k,n)$. We give a criterion for when the realizable classes in the orthogonal Grassmannian agree with those in the Grassmannian. We give a characterization for realizability in low dimension and codimension and give necessary conditions for realizability in certain orthogonal Grassmannians.
Read the full abstract ▾
Geophysics
Climate-informed cryospheric reanalysis via hierarchical Bayesian data assimilation
Kristoffer Aalstad, Esteban Alonso-González, Joel Fiddes et al. (+7 more)
The cryosphere regulates global cycles of water, energy, and carbon, affecting societies and ecosystems. By assimilating observations into models, we can infer the past trajectory of state variables, generating a cryospheric reanalysis. Existing cryospheric reanalyses typically treat each water year independently, with no pooling of information across years. With sparse and noisy observations, such '
climate-assumed' reanalyses perform poorly, reverting to an assumed yet uncalibrated background climatology. Conversely,
Bayesian calibration with static parameters completely pools information across water years, inferring climatological parameters without capturing inter-annual variability. Here, we propose a new hierarchical
Bayesian approach using partial pooling, which we coin '
climate-informed' cryospheric reanalysis. It jointly infers state trajectories, annual parameters, and local climatological hyperparameters using a nested hybrid particle smoothing workflow. We test the approach through three experiments by assimilating in situ snow water equivalent, satellite-based fractional snow-covered area, and glacier-wide mass balance data into a temperature index model at eight study areas, from the Swiss Alps to the High Arctic. Hierarchical reanalysis with partial pooling outperforms annually independent cryospheric reanalysis (no pooling) and static parameter calibration (complete pooling) for all experiments during calibration and for all but one during validation. Overall, it yields a 51% improvement in continuous ranked probability score relative to the prior, markedly exceeding that of annual reanalysis (38%) and static calibration (19%). Compared to a particle Markov chain Monte Carlo
benchmark, our workflow reduces computational cost by several orders of magnitude through recycling and expectation maximization, with comparable or improved skill.
Read the full abstract ▾
All authors Kristoffer Aalstad, Esteban Alonso-González, Joel Fiddes, Brian Groenke, Gregoire Guillet, Regine Hock, Bartłomiej Luks, Norbert Pirk, Sebastian Westermann, Yeliz A. Y{\i}lmaz
Astrophysics of Galaxies
CHAMPS: Data Reduction and Detection of ~3300 Dusty Galaxies at 1.2 mm
Felix Martinez III, Jeyhan S. Karaltepe, Andreas L. Faisst et al. (+52 more)
We present the reduction and source catalogs for the \textit{COSMOS High-$z$ ALMA-MIRI Population Survey} (CHAMPS), the largest Atacama Large (Sub)Millimeter Array (ALMA) blank field survey to date at 0.2\,deg$^2$. With observing frequency centered at $250\, $GHz ($1.2\,$mm) and bandwidth of 7.5 GHz, CHAMPS covers the JWST/NIRCam+MIRI footprint in the COSMOS field, taking advantage of the rich, multiwavelength ancillary
datasets in COSMOS ranging from X-ray to
radio. The majority of the CHAMPS survey (90$\%$) has depths at or below 200 $\mu$Jy beam$^{-1}$ and with an average RMS of 146 $\mu$Jy beam$^{-1}$. The reduced CHAMPS mosaic has a beam size of $\theta_{\rm res} = 1.1'' \times 1.1''$. CHAMPS identifies sources via two detection methods, a blind search for high-SNR sources, and a prior-based search involving MIRI detections to identify faint sources. The blind search yields a total of 595 sources down to an SNR $\ge4.5\sigma$ while the prior catalog yields a total of 3162 source down to an SNR $\ge3\sigma$, leading to a full sample of 3355 unique objects detected across both catalogs. We identify a unique population of 385 SNR $\ge5\sigma$ sources across the two catalogs; after excluding 9 likely spurious edge detections we define a robust science sample of 376 sources above $\ge5\sigma$. These sources have a median
redshift of $\langle z_{1.2 \text{ mm}}\rangle = 2.54^{+0.89}_{-1.69}$ based on optical/near-IR/mid-IR photometric
redshifts and high-quality spectroscopic
redshifts, and hold a median
stellar mass of $\langle\log (\rm M_\star/M_\odot)\rangle = 10.79^{+0.42}_{-3.08}$. We have further identified 68/376 ($\sim18\%$) sources as candidate AGN through X-ray, UV-optical, infrared and
radio selection techniques.
Read the full abstract ▾
All authors Felix Martinez III, Jeyhan S. Karaltepe, Andreas L. Faisst, Arianna S. Long, Jorge A. Zavala, Caitlin M. Casey, Jed McKinney, Ezequiel Treister, Maximilien Franco, Sune Toft, Manuel Aravena, John D. Silverman, Hollis B. Akins, Hiddo S. B. Algera, Jacqueline Antwi-Danso, Andrew J. Battisti, Yingjie Cheng, Olivia R. Cooper, Nicole E. Drakos, Carter Flayhart, Fabrizio Gentile, Steven Gillman, Ghassem Gozaliasl, Santosh Harish, Simon Hempel-Costello, Michaela Hirschmann, Olivier Ilbert, Bingcheng Jin, Shuowen Jin, Koki Kakiichi, Thresa Kelly, Zihao Li, Daizhong Liu, Lun-Jun Liu, Sinclaire M. Manning, Henry Joy McCracken, Romain A. Meyer, Louise Paquereau, Maria Pudoka, Jason Rhodes, Brant E. Robertson, Giulia Rodighiero, Rasha M. Samir, David B. Sanders, Marko Shuntov, Margherita Talia, Mattia Vaccari, Francesco Valentino, Aswin P. Vijayan, Feige Wang, Wuji Wang, Can Xu, Jinyi Yang, Yongda Zhu, Siwei Zou
Computational Physics
Permeability and microcrack geometry: Dynamic loading induced evolution
Rigoberto Moncada and Efrem Vitali
We propose a crack-geometry-based model for permeability
evolution in rocks under dynamic loading. Accurate representation of permeability is important for fluid extraction and containment in geological formations, but remains challenging because low-porosity brittle rocks are highly sensitive to changes in pore geometry and connectivity, while dynamic permeability measurements are limited. Rock void space is represented as an evolving network of penny-shaped cracks with prescribed orientations. Permeability changes are described through crack microvariables including aperture, length, and crack distance. The model accounts for crack opening and closure, fracture-energy-driven propagation, plastic-strain-driven nucleation, coalescence after full connectivity is reached, and strain-rate-dependent fracture toughening. The evolving crack network directly modifies connectivity, fluid conductance, and preferential flow paths. The formulation is implemented in the GeoDyn hydrocode and evaluated under tension, shear, compression, and symmetric impact loading. Results show a strong relation between permeability and crack length, with crack propagation playing a major role in permeability enhancement. Permeability growth can nevertheless be limited by poor connectivity and crack closure. The model captures changes of several orders of magnitude depending on stress state, strain rate, crack orientation, and connectivity, consistent with observed behavior of brittle rocks under high strain-rate loading.
Read the full abstract ▾
Machine Learning
GEM-KMeans: Memory-Efficient and Accurate Clustering on Massive Scale with GPU Optimization
Peng Xu, Nihar Koganti, Volodymyr Kindratenko et al. (+1 more)
Memory-efficient scaling on clustering problems without sacrificing statistical accuracy is of central interest for large-scale data analysis and machine learning problems. Nonnegative low-rank (NLR)
matrix factorization for $K$-means is a scalable clustering method, which connects to semidefinite relaxations with optimal average-case exact recovery guarantees. However, a direct GPU implementation of NLR requires multiple large factor-sized buffers and substantial data movements that are essentially
memory-bound. In this paper, we introduce GEM-KMeans, a spectrally normalized yet mathematically equivalent NLR formulation that fuses the gradient update, nonnegative projection, and sufficient statistics for normalization and iterate movement into a
matrix-multiplication epilogue. Instead of retaining three massive factor-sized arrays, our IO-aware GPU implementation materializes only one single factor with small tile-reduction arrays as additional storage in the High Bandwidth
Memory (HBM). We derive explicit
memory costs and spectrally normalized smoothness bounds for optimizing the clustering objective function. Accurate clustering is demonstrated at massive scales on synthetic and real
datasets, where performance gains of GEM-KMeans over existing GPU-accelerated Lloyd's
algorithms involve data-dependent runtime tradeoffs.
Read the full abstract ▾
All authors Peng Xu, Nihar Koganti, Volodymyr Kindratenko, Xiaohui Chen